add-coder 0.1.13 → 0.1.15

package/README.en.md

@@ -0,0 +1,218 @@
+# add-coder
+
+> 🌐 [🀄中文](./README.md) | English
+
+**AI Governance, Implemented** — The complete scaffolding from [codein2027](https://github.com/xiaomingming92/codein2027) for rapidly building the ADD programming paradigm. Built on the core principle of **Audit as Infrastructure**, it shatters the black-box programming process and cross-session amnesia, evolving the programming paradigm into an auditable, traceable, and convergent new era. [NPM](https://www.npmjs.com/package/add-coder) · [GitHub](https://github.com/xiaomingming92/add-coder)
+
+
+> 🧭 **Getting hands-on?** See [GUIDE.md](https://github.com/xiaomingming92/add-coder/blob/main/GUIDE.md) — trigger word quick reference, requirements-to-Plan, and full workflow walkthrough.
+
+```bash
+npx add-coder init
+```
+
+---
+
+## Not a Template Tool — An Architectural Difference
+
+There are already plenty of AI coding templates, hook adapters, and MCP scaffolds. What fundamentally sets add-coder apart is not "what files it generates," but a **paradigm shift at the architectural level**:
+
+### ① Audit Is Infrastructure, Not After-the-Fact Logging
+
+Traditional AI development: Chat → Generate code → Dig through chat history afterward to find "who changed what"
+
+| Traditional Model | add-coder |
+|-------------------|-----------|
+| Logs are append-only text files | Audit is a **structured data table** (DevOperation + AuditLog), supporting multi-dimensional queries by plan/step/agent/tool |
+| Auditing relies on developer discipline | The **MCP audit toolchain** automatically records every operation; system gateways enforce checks |
+| No traceability | Audit events are naturally linked: Plan → Spec → Task → Step → Tool Call, forming a complete evidence chain |
+
+### ② Gateway-Driven, Not Free-Form Conversation
+
+Traditional AI coding is "you say, I do" — quality depends entirely on the LLM's state that day. add-coder embeds **dual quality gateways** into the architecture:
+
+```
+DPS (Design-Process Symmetry)  — Design / Implementation / Docs / Audit, each weighted 25%, < 85% BLOCKED
+RAHS (Runtime Architecture Health Score) — Runtime architecture health, < 90% BLOCKED
+```
+
+These are not "suggestions" — they are **architectural blocks**. A Step cannot advance without passing its gateway.
+
+### ③ Cross-Session Memory, Not Per-Session Amnesia
+
+The fatal flaw of AI conversations: architectural decisions from last session, bugs fixed, agreements reached — all forgotten in the next conversation. add-coder solves this at the architecture level:
+
+- **Handoff Documents** — Automatically generated structured handoff at the end of each session, auto-loaded by the next session
+- **Plan Index** — All Plans are centrally indexed via `index.md`, supporting fuzzy-match quick lookup
+- **DevLog Timeline** — Every operation is written to the `{YYYY-MM}/{DD}/` timeline, enabling full historical state traceability
+
+### ④ Policy-Update-Loop: Self-Evolving Governance (the scaffold itself does not include this architectural capability; a DEMO repo will be provided next to better illustrate the Policy-Update-Loop and Report system)
+
+Not a static template, but a **closed-loop adaptive system**:
+
+```
+Execute → Audit → Boundary Report → Rule Adjustment → Next Execution
+```
+
+Runtime-generated Reports feed back into governance rules, enabling continuous evolution of governance strategies.
+
+### ⑤ Multi-IDE Hooks as the Governance Layer
+
+Hooks are not "notification push" — they are the **IDE runtime interception layer**:
+
+| Hook Type | Function |
+|-----------|----------|
+| PreToolUse | Validates whitelist before tool invocation, injects context, DPS condition checks |
+| PostToolUse | Automatic audit logging, Plan sync detection, format guarding |
+| PreCompact | Forces retention of critical document paths during cross-session context compression |
+| PromptSubmit | Injects ADD vocabulary triggers, ensuring zero-latency LLM response to commands like "acceptance" and "gateway" |
+
+Each IDE (Claude/Qoder/VS Code) has its own hook implementation, but the **governance logic is unified** — the architecture is consistent, only the adapter layer differs.
+
+---
+
+## Quick Start
+
+```bash
+npx add-coder init
+```
+
+The first `init` auto-detects your IDE and interactively guides you through database selection (PostgreSQL / SQLite / self-managed), container runtime (podman / docker / self-managed), Prisma initialization, and ADD template deployment.
+
+```bash
+npx add-coder init
+# → Choose IDE (Qoder / Claude / VS Code)
+# → Choose database (PostgreSQL / SQLite / self-managed)
+# → Choose container (podman / docker / self-managed)
+# → prisma init + add.prisma copied
+# → prisma db push (adds new tables only, no data deletion)
+# → prisma generate
+# → ADD governance model ready ✓
+```
+
+> **Env file priority**: `.env.development.local` > `.env.development` > `.env.local` > `.env`
+
+## Commands
+
+| Command | Description |
+|---------|-------------|
+| `init` | Initialize ADD templates, supports `--adapter claude\|qoder\|vscode\|auto` |
+| `sync` | Incrementally sync missing files |
+| `status` | Check template integrity |
+
+### init Internal Flow
+
+| Step | Action | Description |
+|------|--------|-------------|
+| ① | Detect IDE | Scan for `.qoder/` `.claude/` `.vscode/` existence, or specify via `--adapter` |
+| ② | Load config | Interactive Q&A > `add-coder.config.ts` > auto-detect > defaults |
+| ③ | DB deployment | `db-ensure.sh` starts container/PG connection + `injectPrisma()` Caijue layer (Prisma init → AddUser model copy → db push → generate) |
+| ④ | Render templates | 55 core template files (skills/agents/templates/plans/specs/scripts…) |
+| ⑤ | Deploy adapters | Copy core content to `.add/` `.qoder/` `.claude/` directories, supplement IDE-specific hooks/mcp |
+| ⑥ | Write files | Four modes: interactive / yes / force / dry-run; `.sh` scripts auto `chmod` |
+| ⑦ | Output summary | Created / skipped / overwritten stats + next-step hints |
+
+### init Options
+
+| Option | Description |
+|--------|-------------|
+| `--adapter <type>` | Target IDE: claude / qoder / vscode / auto (default) |
+| `--config <path>` | Specify config file |
+| `--yes` | Skip interactions, create new files only |
+| `--force` | Overwrite existing files |
+| `--dry-run` | Preview mode, no writes |
+
+## Generated Content
+
+| Directory | Content |
+|-----------|---------|
+| `.add/` | ADD shared core (skills, agents, docs, scripts, rules, etc.) |
+| `.claude/` | Claude Code adapter (hooks, settings.json) |
+| `.qoder/` | Qoder adapter (hooks, settings.json, mcp.json) |
+| `.vscode/` | VS Code adapter (settings.json, tasks.json) |
+
+## MCP Audit Toolchain
+
+`init` automatically deploys the MCP server (`mcp-server.ts`) into the project, loaded by the IDE via `mcp.json`. The following audit and governance tools are provided:
+
+| Tool | Purpose | Trigger Scenario |
+|------|---------|-----------------|
+| `record_dev_operation` | Record development operation audits (before/after/reason) | Every file change, config modification |
+| `query_audit_logs` | Query audit records by planKeyword / targetId | Cross-session context recovery, iteration evidence verification |
+| `get_project_context` | Get ADD workflow status snapshot | Fresh conversation start |
+| `get_db_schema` | Get Prisma schema info | Database-related operations |
+| `check_dps` | DPS gateway (Design/Implementation/Docs/Audit, each 25%) | End of Step 0 |
+| `check_rahs` | RAHS gateway (runtime architecture health) | Step 4/8 |
+| `check_add_route_status` | add-route file existence check | Before Step 3 |
+| `check_spec_sync` | Spec doc checkbox status vs. code consistency | After Spec execution |
+| `find_related_docs` | Search related architecture/spec documents | Context understanding |
+
+> Full tool list: [MCP Toolchain Specification](https://github.com/xiaomingming92/codein2027/blob/main/docs/大田精准耕播智能决策系统/knowledge/02-规范/%E3%80%8A%E5%BC%80%E5%8F%91%E6%93%8D%E4%BD%9C%E5%AE%A1%E8%AE%A1%E5%AD%98%E6%A1%A3%E8%A7%84%E8%8C%83%E3%80%8B.md).
+
+## Architecture Overview
+
+```
+                    ┌─────────────┐
+                    │  ADD Paradigm│
+                    │  Step 0-9    │
+                    └──────┬──────┘
+                           │ Gateway-driven
+              ┌────────────┼────────────┐
+              ▼            ▼            ▼
+       ┌──────────┐ ┌──────────┐ ┌──────────┐
+       │DPS Gateway│ │RAHS Gate │ │Compliance│
+       └────┬─────┘ └────┬─────┘ └────┬─────┘
+            │            │            │
+              ▼            ▼            ▼
+       ┌─────────────────────────────────────┐
+       │        Audit Infrastructure Layer    │
+       │  DevOperation / AuditLog Tables      │
+       │  MCP Audit Toolchain                 │
+       │  Handoff / DevLog Timeline Docs      │
+       └─────────────────────────────────────┘
+                           │
+              ┌────────────┼────────────┐
+              ▼            ▼            ▼
+       ┌──────────┐ ┌──────────┐ ┌──────────┐
+       │ Claude   │ │  Qoder   │ │  VS Code │
+       │ Hooks    │ │  Hooks   │ │  Config  │
+       └──────────┘ └──────────┘ └──────────┘
+                           │
+              ┌────────────┘
+              ▼
+       ┌─────────────────────────────────────┐
+       │       Caijuehub Rule Engine          │
+       │  TOML-Driven Policy System           │
+       │  Detect / Adapt / Prisma / Write —   │
+       │  Fully Configurable                  │
+       └─────────────────────────────────────┘
+```
+
+## Prerequisites
+
+- Node.js >= 20
+- Prisma ^7.0 (auto-detected during `init`, guided installation if missing)
+- PostgreSQL / SQLite (MCP toolchain depends on DevOperation + AuditLog tables)
+
+> **Recommended**: Run PostgreSQL via Podman/Docker, reference config:
+> ```yaml
+> postgres:
+>   image: docker.io/postgres:16-alpine
+>   ports: ["127.0.0.1:5433:5432"]
+>   environment:
+>     POSTGRES_DB: mydb
+>     POSTGRES_USER: admin
+>     POSTGRES_PASSWORD: ${POSTGRES_PASSWORD}
+> ```
+> Mount data volume to `~/data/your_project/postgres/` to avoid data loss on container removal.
+
+> 📦 [Changelog](./CHANGELOG.md)
+
+---
+
+## 🎬 Coming Soon
+
+| Plan | Description |
+|------|-------------|
+| Demo Repo | A full example repository showcasing end-to-end closed-loop practice of Policy-Update-Loop and the Report system |
+| MCP Restructure | MCP toolchain architecture upgrade, improving audit and gateway tool extensibility and standalone deployment capability |

package/README.md

@@ -1,7 +1,10 @@
 # add-coder
 
+> 🌐 🀄中文 | [English](./README.en.md)
+
 **AI 代码治理的落地方案** — [codein2027](https://github.com/xiaomingming92/codein2027) 快速构建 ADD 编程范式的完整脚手架。以「审计即基础设施」为核心，彻底打破编程过程黑盒与跨轮失忆，让编程范式进化为可审计、可追溯、可收敛的新时代。 [NPM](https://www.npmjs.com/package/add-coder) · [GitHub](https://github.com/xiaomingming92/add-coder)
 
+
 > 🧭 **从零上手实操？** 请参见 [GUIDE.md](https://github.com/xiaomingming92/add-coder/blob/main/GUIDE.md) — 包含触发词速查、需求转 Plan、完整链路演练。
 
 ```bash
@@ -43,7 +46,7 @@ AI 对话的致命缺陷：上次讨论的架构决策、已修复的 Bug、达
 - **Plan 索引** — 所有 Plan 通过 `index.md` 集中索引，支持模糊匹配快速定位
 - **DevLog 时序记录** — 每一步操作写入 `{YYYY-MM}/{DD}/` 时间轴，可回溯任意历史状态
 
-### ④ Policy-Update-Loop：治理自我进化
+### ④ Policy-Update-Loop：治理自我进化(脚手架不包含此架构能力,接下来会给到DEMO仓库让大家更好理解Policy-Update-Loop和Report体系)
 
 不是静态模板，而是**闭环自适应系统**：
 
@@ -201,3 +204,14 @@ npx add-coder init
 >     POSTGRES_PASSWORD: ${POSTGRES_PASSWORD}
 > ```
 > 数据卷建议挂载到 `~/data/your_project/postgres/`，避免容器销毁丢失数据。
+
+> 📦 [更新日志 (CHANGELOG)](./CHANGELOG.md)
+
+---
+
+## 🎬 预告
+
+| 计划 | 说明 |
+|------|------|
+| Demo 仓库演示 | 提供完整示例仓库，展示 Policy-Update-Loop 与 Report 体系的端到端闭环实践 |
+| MCP 能力重构 | MCP 工具链架构升级，提升审计与门禁工具的可扩展性和独立部署能力 |

package/dist/index.js

@@ -759,7 +759,7 @@ PROJECT_NAME=${config.projectName || "add-project"}
   }
   const coreFiles = renderCore(config, !!options.dryRun);
   console.log(`Core \u6A21\u677F: ${coreFiles.size} \u6587\u4EF6`);
-  const CORE_TARGETS = [".add", ".qoder", ".claude", ".vscode"];
+  const CORE_TARGETS = [".add", magicDir];
   const allFiles = /* @__PURE__ */ new Map();
   for (const [relPath, content] of coreFiles) {
     for (const t of CORE_TARGETS) {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "add-coder",
-  "version": "0.1.13",
+  "version": "0.1.15",
   "private": false,
   "description": "[codein2027](https://github.com/xiaomingming92/codein2027) 快速构建 ADD 编程范式的完整脚手架——AI 代码治理的落地方案。以「审计即基础设施」为核心，彻底打破编程过程黑盒与跨轮失忆，让编程范式进化为可审计、可追溯、可收敛的新时代。npx 即用，人人可体验。",
   "type": "module",

package/templates/core/hooks/doc-format-guard.sh

@@ -0,0 +1,17 @@
+#!/bin/bash
+# doc-format-guard.sh — 文档格式守卫（Claude Code 适配）
+set -euo pipefail
+
+HOOK_DIR="$(cd "$(dirname "$0")" && pwd)"
+SHARED_LIB="$HOOK_DIR/../../shared/hooks-lib/common.sh"
+[ -f "$SHARED_LIB" ] && source "$SHARED_LIB"
+
+input=$(parse_input)
+file_path=$(echo "$input" | jq -r '.tool_input.file_path // empty')
+[ -z "$file_path" ] && exit 0
+
+# 检查 ADD 文档格式
+if echo "$file_path" | grep -qE '\.qoder/(specs|plans|reviews)/.*\.md$'; then
+  echo "[ADD DocFormatGuard] 检测到 ADD 文档变更，请确保格式符合模板规范。"
+fi
+exit 0

package/templates/core/hooks/lib/context-inject.sh

@@ -0,0 +1,96 @@
+#!/bin/bash
+# context-inject.sh — 上下文注入模板
+# 共享库
+
+# 结构化 SessionStart JSON 注入
+build_session_start_json() {
+  local plan="$1" step="$2" round="$3" handoff="$4"
+  cat <<EOF
+{
+  "continue": true,
+  "hookSpecificOutput": {
+    "hookEventName": "SessionStart",
+    "additionalContext": "上次 ADD 流程未完成:\n  Plan: ${plan}\n  轮次: ${round}\n  当前 Step: ${step} (add-route)\n  恢复命令: query_audit_logs({ planKeyword: '${plan}' })\n  handoff: ${handoff}"
+  }
+}
+EOF
+}
+
+# Stop 四象限分流 few-shot 上下文
+build_stop_context() {
+  local quadrant="$1" info="$2"
+  case "$quadrant" in
+    "no_add_no_dev")
+      echo "[ADD Stop] 无活跃 Plan，无代码改动。正常结束。"
+      ;;
+    "no_add_has_dev")
+      cat <<'EOF'
+[ADD Stop] ⚠️ 检测到代码修改但无活跃 ADD Plan。
+
+Plan 不是"文档开销"——它是代码治理的基础设斻。跳过 Plan 的后果:
+  · 没有 add-route → 每次改动无法追溯到具体 Task
+  · 没有 tasks.md → 后续 AI Session 不知道改了哪些文件
+  · 没有 handoff → 交接时上下文全丢，只能靠 git log 猜
+
+你必须立即补救，二选一:
+
+方案 A — 补 ADD 流程（招安）:
+  Step 0: 读 .qoder/templates/plan-template.md → 生成 Plan → .qoder/plans/{today}/{keyword}-plan-v1.md
+          生成 add-route → check_dps ≥ 85
+  Step 1: 扩展 AgentAuditPhase（如需要）
+  Step 2: 确认 agentAudit() 通道
+  Step 3: 将已写代码关联到 tasks.md
+  完成后可正常停止。
+
+方案 B — 补不上则回滚:
+  如果改动太复杂无法追溯生成 Plan，则:
+  ① git diff 确认改动范围
+  ② git checkout -- <改动的文件> 撤回
+  ③ 告知用户 "以下改动已回滚，请先走 ADD 流程再重新实施: <文件列表>"
+
+无论选 A 还是 B，完成后告诉用户下次执行 session-init 恢复上下文。
+EOF
+      ;;
+    "has_add_no_dev")
+      echo "[ADD Stop] ADD 流程进行中 (${info})，本次无代码改动。下次继续时执行 session-init 恢复上下文。"
+      ;;
+    "has_add_dev_step02")
+      echo "[ADD Stop] ADD Step 0-2: 文档先行/审计准备阶段。无需验收闭环。下一步: 进入 Step 3 代码实现。"
+      ;;
+    "has_add_dev_step3")
+      echo "[ADD Stop] ADD Step 3: 代码实现进行中 (${info})。完成后进入 Step 3.5 实现审查。"
+      ;;
+    "has_add_dev_unclosed")
+      cat <<'EOF'
+[ADD Stop] ⚠️ 代码已完成但验收未闭环:
+${info}
+
+请依次执行（不要等下次会话）:
+  ① Write devlog → handoff 同目录 devlog-{plan}-v{n}.md
+     格式: # Devlog: {plan}\n 日期 / Plan / 轮次 / 本轮改了什么 / 验收结果 / 遗留项 / 架构回看
+  ② Edit handoff → 更新 §验证标准 全部 [x] + 补充审计 ID
+     ★ 同步: checklist 有新 cuid → handoff ADD-7 表必须对应新增行
+     ★ Step 0 准入: handoff + add-route + Specs 三元组缺一不可，缺则回退 Step 0.5
+  ③ Read docs/ → 回看架构文档确认一致性
+  ④ Edit add-route → 勾选对应 Step [x]
+
+以上全部完成后 Agent 才能停止。
+
+下次恢复: 读 handoff → 查同目录 devlog-*.md → query_audit_logs
+EOF
+      ;;
+        "has_add_dev_closed")
+      echo "[ADD Stop] ✅ 验收闭环: add-route全部[x], devlog已记录, handoff已更新。验收幂等——重复触发不覆盖已有结论。"
+      ;;
+  esac
+}
+
+# 写操作前置守卫上下文
+build_pretool_context() {
+  local plan="$1" round="$2"
+  cat <<EOF
+[ADD PreToolUse] 当前 Plan: ${plan}，轮次: ${round}。
+本次写入应属于 ADD Step 3 代码实现阶段。
+完成后执行 record_dev_operation 记录审计。
+EOF
+}

package/templates/core/hooks/lib/state-detect.sh

@@ -0,0 +1,104 @@
+#!/bin/bash
+# state-detect.sh — ADD 活跃流程检测 + dev action 追踪
+# 共享库
+
+PROJECT_DIR="${QODER_PROJECT_DIR:-${QODERCN_PROJECT_DIR:-$PWD}}"
+QODER_DIR="$PROJECT_DIR/.qoder"
+PLANS_DIR="$QODER_DIR/plans"
+
+# dev action 标记文件（项目级，PreToolUse 写入，Stop 读取）
+DEV_FLAG="/tmp/qoder_dev_$(echo "$PROJECT_DIR" | md5sum 2>/dev/null | cut -c1-8 || echo "default")"
+
+# 检测活跃 ADD 流程
+# 返回: "plan_keyword::step_x/total::round_n/total_r::handoff_path::add_route_path" 或 ""
+detect_active_add() {
+  local handoff=""
+  local today_dir="$PLANS_DIR/$(date +%Y-%m)"
+  [ -d "$today_dir" ] && handoff=$(find "$today_dir" -name "*handoff*.md" -mtime -7 -type f 2>/dev/null | head -1)
+  [ -z "$handoff" ] && handoff=$(find "$PLANS_DIR" -name "*handoff*.md" -mtime -7 -type f 2>/dev/null | head -1)
+  [ -z "$handoff" ] && return 1
+
+  local plan_kw=$(basename "$handoff" | sed 's/-handoff.*//')
+
+  # add-route
+  local add_route=$(find "$PLANS_DIR" -name "*add-route*.md" -type f 2>/dev/null | head -1)
+
+  # Step 统计（从 add-route）
+  local step_info="?"
+  if [ -f "$add_route" ]; then
+    local ck=$(grep -c '\[x\]' "$add_route" 2>/dev/null || true)
+    local uc=$(grep -c '\[ \]' "$add_route" 2>/dev/null || true)
+    ck=${ck:-0}; uc=${uc:-0}
+    local total=$((ck + uc))
+    [ "$total" -gt 0 ] && step_info="${ck}/${total}"
+  fi
+
+  # 轮次统计（从 handoff）
+  local round_info="?"
+  if grep -q '<第[0-9]轮>' "$handoff" 2>/dev/null; then
+    local ck=$(grep -c '\[x\]' "$handoff" 2>/dev/null || true); ck=${ck:-0}
+    local total=$(grep -c '<第[0-9]轮>' "$handoff" 2>/dev/null || true); total=${total:-0}
+    [ "$total" -gt 0 ] && round_info="${ck}/${total}"
+  else
+    round_info="1/1"
+  fi
+
+  echo "${plan_kw}::${step_info}::${round_info}::${handoff}::${add_route:-none}"
+}
+
+# 标记 dev action
+mark_dev_action() {
+  touch "$DEV_FLAG" 2>/dev/null || true
+}
+
+# 检测 dev action
+has_dev_action() {
+  [ -f "$DEV_FLAG" ]
+}
+
+# 清理 dev action 标记（验收闭环后调用）
+clear_dev_action() {
+  rm -f "$DEV_FLAG" 2>/dev/null || true
+}
+
+# 检查验收完整度，返回 issues 文本
+check_add_completeness() {
+  local handoff="$1" add_route="$2"
+  local issues=""
+
+  # devlog（内容已回流至 handoff，检查 handoff 是否含验收结果）
+  if [ -f "$handoff" ] && ! grep -qE '验收|收敛|闭环|本轮改了什么|devlog' "$handoff" 2>/dev/null; then
+    issues="${issues}  [ ] devlog 缺失（handoff 无验收记录）\n"
+  fi
+
+  # handoff 验证标准
+  if [ -f "$handoff" ]; then
+    local uc=$(grep -c '\[ \]' "$handoff" 2>/dev/null || echo "0")
+    [ "$uc" -gt 0 ] && issues="${issues}  [ ] handoff ${uc} 项未勾选\n"
+  fi
+
+  # add-route Step
+  if [ -f "$add_route" ]; then
+    local uc=$(grep -c '\[ \]' "$add_route" 2>/dev/null || echo "0")
+    [ "$uc" -gt 0 ] && issues="${issues}  [ ] add-route ${uc} Step 未闭环\n"
+  fi
+
+  echo -e "$issues"
+}
+
+# 检查是否已验收（幂等保护）
+# 返回 0 = 未验收，1 = 已验收
+is_already_accepted() {
+  local add_route="$1" handoff="$2"
+  # add-route Step 8 是否已 [x]
+  if [ -f "$add_route" ]; then
+    # 检查 Step 8 产出项是否勾选
+    if grep -A 10 'Step 8' "$add_route" 2>/dev/null | grep -q '\[x\].*验证并更新项目状态'; then
+      # handoff 是否有验收记录
+      if [ -f "$handoff" ] && grep -qE '✅.*验收|收敛|全部闭环|全部.*完成' "$handoff" 2>/dev/null; then
+        return 0
+      fi
+    fi
+  fi
+  return 1
+}

package/templates/core/hooks/lib/vocabulary.sh

@@ -0,0 +1,49 @@
+#!/bin/bash
+# vocabulary.sh — 从 vocabulary markdown 表格加载触发词
+# 单一数据源: .qoder/vocabulary/add-governance-vocabulary.md §类别 A-F 表格
+
+VOCABULARY_FILE="$PWD/.qoder/vocabulary/add-governance-vocabulary.md"
+
+# 输出格式: 优先级::触发词正则::响应文本（:: 避免与触发词内的 | 冲突）
+load_triggers() {
+  [ ! -f "$VOCABULARY_FILE" ] && return 1
+  sed -n '/^## 类别 A: 文档类型/,/^## 类别 [G-Z]/p' "$VOCABULARY_FILE" \
+    | grep -E '^\| (P0|P1|P2) ' \
+    | while IFS='|' read -r _ prio raw_trigger action; do
+        prio=$(echo "$prio" | xargs)
+        trigger=$(echo "$raw_trigger" | sed 's/`//g;s/ *\/ */|/g;s/^ *//;s/ *$//')
+        action=$(echo "$action" | xargs)
+        [ -z "$trigger" ] && continue
+        echo "${prio}::${trigger}::${action}"
+      done || true
+}
+
+match_trigger() {
+  local prompt="$1"
+  while IFS= read -r line; do
+    [ -z "$line" ] && continue
+    local prio="${line%%::*}"
+    local rest="${line#*::}"
+    local regex="${rest%%::*}"
+    local action="${rest#*::}"
+    # 跳过开发关键词检测行（由 Layer 2/3 分流处理，特征：超长regex含"修bug"）
+    if echo "$regex" | grep -qE '修\.\?bug|fix\.\?bug' 2>/dev/null; then
+      continue
+    fi
+    if echo "$prompt" | grep -qiE "$regex" 2>/dev/null; then
+      echo "[ADD 触发] ${regex} → ${action}"
+    fi
+  done < <(load_triggers 2>/dev/null)
+}
+
+load_dev_keywords() {
+  load_triggers 2>/dev/null | while IFS= read -r line; do
+    [ -z "$line" ] && continue
+    # 只取包含"开发|改功能"的开发关键词检测行
+    if echo "$line" | grep -q '开发|改功能' 2>/dev/null; then
+      local rest="${line#*::}"
+      local regex="${rest%%::*}"
+      echo "$regex"
+    fi
+  done || true
+}

package/templates/core/hooks/notification.sh

@@ -0,0 +1,9 @@
+#!/bin/bash
+# Notification — 通知事件处理（Claude Code 适配）
+set -euo pipefail
+
+HOOK_DIR="$(cd "$(dirname "$0")" && pwd)"
+SHARED_LIB="$HOOK_DIR/../../shared/hooks-lib/common.sh"
+[ -f "$SHARED_LIB" ] && source "$SHARED_LIB"
+
+exit 0

package/templates/core/hooks/permission-gate.sh

@@ -0,0 +1,18 @@
+#!/bin/bash
+# PermissionRequest — 权限请求门禁（Claude Code 适配）
+set -euo pipefail
+
+HOOK_DIR="$(cd "$(dirname "$0")" && pwd)"
+SHARED_LIB="$HOOK_DIR/../../shared/hooks-lib/common.sh"
+[ -f "$SHARED_LIB" ] && source "$SHARED_LIB"
+
+input=$(parse_input)
+tool_name=$(echo "$input" | jq -r '.tool_name // empty')
+
+# 高风险工具需要二次确认
+case "$tool_name" in
+  Bash|Write|Edit)
+    echo "[ADD PermissionGate] 高风险工具: ${tool_name}，请确认操作。"
+    ;;
+esac
+exit 0

package/templates/core/hooks/post-tool-failure.sh

@@ -0,0 +1,10 @@
+#!/bin/bash
+# PostToolUseFailure — 工具失败后处理（Claude Code 适配）
+set -euo pipefail
+
+HOOK_DIR="$(cd "$(dirname "$0")" && pwd)"
+SHARED_LIB="$HOOK_DIR/../../shared/hooks-lib/common.sh"
+[ -f "$SHARED_LIB" ] && source "$SHARED_LIB"
+
+echo "[ADD PostToolFailure] 工具调用失败，请检查错误信息并修复。"
+exit 0

package/templates/core/hooks/post-tool-use.sh

@@ -0,0 +1,18 @@
+#!/bin/bash
+# post-tool-use.sh — Claude Code PostToolUse 审计提醒
+set -euo pipefail
+
+HOOK_DIR="$(cd "$(dirname "$0")" && pwd)"
+SHARED_LIB="$HOOK_DIR/../../shared/hooks-lib/common.sh"
+[ -f "$SHARED_LIB" ] && source "$SHARED_LIB"
+
+input=$(parse_input)
+file_path=$(echo "$input" | jq -r '.tool_input.file_path // empty')
+[ -z "$file_path" ] && exit 0
+
+if ! echo "$file_path" | grep -qE 'src/agents/|src/lib/agent-audit-logger\.ts|src/types/|prisma/schema\.prisma'; then
+  exit 0
+fi
+
+echo "[ADD PostToolUse] 文件已写入: ${file_path}。如果尚未记录，请执行 record_dev_operation 落库审计。"
+exit 0

package/templates/core/hooks/pre-compact.sh

@@ -0,0 +1,14 @@
+#!/bin/bash
+# PreCompact — 上下文压缩前检查（Claude Code 适配）
+set -euo pipefail
+
+HOOK_DIR="$(cd "$(dirname "$0")" && pwd)"
+SHARED_LIB="$HOOK_DIR/../../shared/hooks-lib/common.sh"
+[ -f "$SHARED_LIB" ] && source "$SHARED_LIB"
+
+# 检查是否有未完成的 ADD 审计记录
+PROJECT_DIR="${CLAUDE_PROJECT_DIR:-$PWD}"
+[ -d "$PROJECT_DIR/.qoder/plans" ] || exit 0
+
+echo "[ADD PreCompact] 上下文压缩前，请确认 ADD-7 审计记录已落库。"
+exit 0

package/templates/core/hooks/pre-tool-use.sh

@@ -0,0 +1,28 @@
+#!/bin/bash
+# pre-tool-use.sh — Claude Code PreToolUse：源码 Plan 关联检查
+# 与 Qoder 版本共享核心逻辑，差异仅在工具名和环境变量
+set -euo pipefail
+
+HOOK_DIR="$(cd "$(dirname "$0")" && pwd)"
+SHARED_LIB="$HOOK_DIR/../../shared/hooks-lib/common.sh"
+[ -f "$SHARED_LIB" ] && source "$SHARED_LIB"
+
+input=$(parse_input)
+file_path=$(echo "$input" | jq -r '.tool_input.file_path // empty')
+[ -z "$file_path" ] && exit 0
+
+if ! echo "$file_path" | grep -qE '(src/|/src/).*\.(ts|tsx)$'; then
+  exit 0
+fi
+
+PROJECT_DIR="${CLAUDE_PROJECT_DIR:-$PWD}"
+MOD=$(basename "$file_path" | sed 's/\.[jt]sx\?$//')
+
+MATCHES=$(grep -rl "$MOD" "$PROJECT_DIR/.qoder/plans" "$PROJECT_DIR/.qoder/specs" "$PROJECT_DIR/.qoder/reports" 2>/dev/null | wc -l)
+
+if [ "$MATCHES" -gt 0 ]; then
+  exit 0
+fi
+
+echo "⛔ 阻断: ${file_path} 无相关 ADD Plan。请先创建 Plan。" >&2
+exit $EXIT_BLOCK

package/templates/core/hooks/prompt-submit.sh

@@ -0,0 +1,16 @@
+#!/bin/bash
+# UserPromptSubmit — 用户输入提交前检查（ADD 关键词兜底）
+set -euo pipefail
+
+HOOK_DIR="$(cd "$(dirname "$0")" && pwd)"
+SHARED_LIB="$HOOK_DIR/../../shared/hooks-lib/common.sh"
+[ -f "$SHARED_LIB" ] && source "$SHARED_LIB"
+
+input=$(parse_input)
+prompt=$(echo "$input" | jq -r '.prompt // empty')
+
+# 检测 ADD 关键词触发
+if echo "$prompt" | grep -qiE '开发|改功能|修.?bug|加需求|新增|重构|实施|验收|继续'; then
+  echo "[ADD PromptSubmit] 检测到开发关键词，请确认已加载 ADD 工作流上下文。"
+fi
+exit 0

package/templates/core/hooks/review-checklist.sh

@@ -0,0 +1,10 @@
+#!/bin/bash
+# review-checklist.sh — Review 检查清单校验（Claude Code 适配）
+set -euo pipefail
+
+HOOK_DIR="$(cd "$(dirname "$0")" && pwd)"
+SHARED_LIB="$HOOK_DIR/../../shared/hooks-lib/common.sh"
+[ -f "$SHARED_LIB" ] && source "$SHARED_LIB"
+
+echo "[ADD ReviewChecklist] 请在提交前确认 checklist.md 全部项已勾选。"
+exit 0

package/templates/core/hooks/session-start.sh

@@ -0,0 +1,23 @@
+#!/bin/bash
+# SessionStart — ADD 上下文恢复（Claude Code 适配）
+set -euo pipefail
+
+HOOK_DIR="$(cd "$(dirname "$0")" && pwd)"
+SHARED_LIB="$HOOK_DIR/../../shared/hooks-lib/common.sh"
+[ -f "$SHARED_LIB" ] && source "$SHARED_LIB"
+
+# 尝试加载 Qoder 的 state-detect 和 context-inject 库
+QODER_LIB="$HOOK_DIR/../../qoder/hooks/lib"
+[ -f "$QODER_LIB/state-detect.sh" ] && source "$QODER_LIB/state-detect.sh"
+[ -f "$QODER_LIB/context-inject.sh" ] && source "$QODER_LIB/context-inject.sh"
+
+export QODER_PROJECT_DIR="${CLAUDE_PROJECT_DIR:-$PWD}"
+
+state=$(detect_active_add 2>/dev/null || true)
+if [ -z "$state" ]; then
+  exit 0
+fi
+
+IFS='::' read -r plan step rounds handoff <<< "$state"
+build_session_start_json "$plan" "$step" "$rounds" "$handoff"
+exit 0

package/templates/core/hooks/stop-check.sh

@@ -0,0 +1,10 @@
+#!/bin/bash
+# Stop — 停止前合规检查（Claude Code 适配）
+set -euo pipefail
+
+HOOK_DIR="$(cd "$(dirname "$0")" && pwd)"
+SHARED_LIB="$HOOK_DIR/../../shared/hooks-lib/common.sh"
+[ -f "$SHARED_LIB" ] && source "$SHARED_LIB"
+
+echo "[ADD Stop] 会话结束前，请确认所有 checklist 项已勾选、ADD-7 审计已落库。"
+exit 0

package/templates/core/hooks/subagent-guard.sh

@@ -0,0 +1,15 @@
+#!/bin/bash
+# SubagentStart/SubagentStop — 子代理门禁（Claude Code 适配）
+set -euo pipefail
+
+HOOK_DIR="$(cd "$(dirname "$0")" && pwd)"
+SHARED_LIB="$HOOK_DIR/../../shared/hooks-lib/common.sh"
+[ -f "$SHARED_LIB" ] && source "$SHARED_LIB"
+
+input=$(parse_input)
+event=$(echo "$input" | jq -r '.hook_event_name // empty')
+
+if [ "$event" = "SubagentStart" ]; then
+  echo "[ADD SubagentGuard] 子代理启动，请确认其遵循 ADD 规范。"
+fi
+exit 0